For some metrics, is difficult to say when it's good or bad... Bounce rate, for example, maybe is good or bad, it depends by the page. Searches metrics, maybe is good when is going down, meaning people is using less the search engine to find content...

Maybe can be useful having a switch in each widget, declaring when we want green for rising metrics and when we want green for lowering metrics?

There is an event Metrics.isLowerValueBetter, which can be used to define that. In some cases that might not be done correctly.

Note: the page generation time will be deprecated in Matomo 4 and replaced with more specific metrics. See #15736

Might be good to quickly look through all metrics and check if there are any other metrics where lower metrics should be treated as better.

Another wrote an email about this bug.

@enual Which metric did the mention? We won't apply further changes to the page generation time, as the metric is deprecated anyway. For another metric we applied a fix in Matomo 5 #20788

Maybe this issue opened for 4.x was addressing unconsciously to that missing fix, now merged into 5.x branch.
AVG. Generation Time, maybe now it's deprecated, but site-search count stills there.

@tassoman I wouldn't count a lower value of site-search count as better in general. There might be sites, that want users to use the search, while for other pages using the search more likely means, that the navigation is bad and users aren't able to directly find what they are looking for. Or did you mean something else?

I wanted to mean your second example, most of times, using "search engine" UX model is a shortcut to avoid bad navigation.
More on this, site-searches are measurable if related to founded results.
While measuring fastest time it's pretty straightforward if it's green or red., there are some metrics are grey-matter.
For example, bounces, site-searches ... etc ...

I'll reclassify this as an enhancement to cater for the 'searches' metric in terms of considering a configuration to determine whether it should be treated as better or worse. Potentially we can look at some more generic approach to these as it's a valid view that some users may see the upwards/downward trends differently.

At the moment I don't think there's a bug here.